Characterization of ERG , AR and PTEN Gene Status in Circulating Tumor Cells from Patients with Castration-Resistant Prostate Cancer

Hormone-driven expression of the ERG oncogene after fusion with TMPRSS2 occurs in 30% to 70% of therapy-naive prostate cancers. Its relevance in castration-resistant prostate cancer (CRPC) remains controversial as ERG is not expressed in some TMPRSS2-ERG androgen-independent xenograft models. However, unlike these models, CRPC patients have an increasing prostate-specific antigen, indicating active androgen receptor signaling. Here, we collected blood every month from 89 patients (54 chemotherapy-naive patients and 35 docetaxel-treated patients) treated in phase I/phase II clinical trials of an orally available, highly specific CYP17 inhibitor, abiraterone acetate, that ablates the synthesis of androgens and estrogens that drive TMPRSS2-ERG fusions. We isolated circulating tumor cells (CTC) by anti–epithelial cell adhesion molecule immunomagnetic selection followed by cytokeratin and CD45 immunofluorescence and 4 ¶ ,6-diamidino-2-phenylindole staining. We used multicolor fluorescence in situ hybridization to show that CRPC CTCs, metastases, and prostate tissue invariably had the same ERG gene status as therapy-naive tumors ( n = 31). We then used quantitative reverse transcription–PCR to show that ERG expression was maintained in CRPC. We also observed homogeneity in ERG gene rearrangement status in CTCs ( n = 48) in contrast to significant heterogeneity of AR copy number gain and PTEN loss, suggesting that rearrangement of ERG may be an earlier event in prostate carcinogenesis. We finally report a significant association between ERG rearrangements in therapy-naive tumors, CRPCs, and CTCs and magnitude of prostate-specific antigen decline ( P = 0.007) in CRPC patients treated with abiraterone acetate. These data confirm that CTCs are malignant in origin and indicate that hormone-regulated expression of ERG persists in CRPC.


Introduction
Hormone-driven expression of the ERG oncogene after fusion with TMPRSS2 occurs in 30% to 70% of therapy-naive prostate cancers (1).Its relevance in castration-resistant prostate cancer (CRPC) remains controversial, as ERG is not expressed in some TMPRSS2-ERG androgen-independent xenograft models (2).However, unlike these models, CRPC patients have an increasing prostate-specific antigen (PSA) indicating active androgen receptor (AR) signaling.Recent studies indicate that the oral and highly specific CYP17 inhibitor, abiraterone acetate, ablates the synthesis of androgens and estrogens that drive TMPRSS2-ERG fusions (1,3), inducing disease regression in >50% of CRPC patients (4).We hypothesized that hormone-dependent overexpression of ERG persisted in CRPC and that TMPRSS2-ERG tumors represented a subgroup of prostate cancers that remained sensitive to CYP17 blockade.We also hypothesized that two potential mechanisms of resistance to CYP17 were gain of AR and loss of PTEN that could result in constitutive phosphorylation of the AR, leading to ligandless activation (5)(6)(7).
We have used anti-epithelial cell adhesion molecule (EpCAM) antibody-based immunomagnetic selection to isolate CRPC circulating tumor cells (CTC; ref. 8) on which to perform multicolor fluorescence in situ hybridization (FISH) investigating the ERG, AR , and PTEN gene loci.As considerable differences exist in how a CTC is defined, with several groups reporting these events differently (9)(10)(11), we describe all anti-EpCAM and cytokeratin (CK)-positive categories.

Materials and Methods
Patients Blood (7.5 mL) was collected into a CellSave tube (Immunicon) from patients with biochemically or histologically confirmed prostate cancer and an increasing PSA despite castrate levels of testosterone (<20 ng/dL) before starting and at monthly intervals until progression on clinical trials of abiraterone acetate.CTCs were enumerated by the CellSearch System (Veridex LLC; ref. 8).Archival tumor acquired before administration of any hormone treatment and blood for CTC analysis were prospectively collected, and patients with untreated primary tumors or accessible metastatic disease were given the option to undergo a tumor biopsy for research purposes.Samples were anonymized by the study trial coordinators, and operators were blinded to clinical outcome.This study was approved by the Royal Marsden NHS Hospital Ethics Review Committees, and informed consent was obtained for these studies.
FISH.Four-color FISH using an ERG break-apart assay (12) and probes covering the AR and the minimal region of deletion at the PTEN gene locus (Supplementary Fig. S1; ref. 13) was performed on CTCs fixed in cartridges with 3:1 methanol/acetic acid.Briefly, CTCs were hybridized overnight at 42jC with the hybridization mix containing rhodamin-labeled PTEN probes, Dy647-labeled 3 ¶-ERG, F415-labeled 5 ¶-ERG, and FITC-labeled AR.Cartridges were washed twice with 24% formamide/0.1ÂSSC at 42jC and then filled with PBS.Cells were then individually captured with a 40Â objective in five Z-stack images over 8 Am for each filter (14).Flattened images were scored for whether they had loss of half or all copies of PTEN (heterozygous or homozygous loss) and increased AR copy number and, for ERG probes, whether they were (a) class ''Normal, '' with unrearranged ERG loci characterized by twinned Dy415 (in red; 3 ¶-ERG) and Dy647 (in green; 5 ¶-ERG) FISH signals; (b) class ''Esplit, '' with a rearranged ERG locus manifested as separation of the red and green signals; or (c) class ''Edel'' ( for ERG 5 ¶-deletion), an ERG rearrangement with loss of green signal corresponding to 5 ¶-ERG sequences.Within classes Edel and Esplit, the pattern could be further divided into cancers that contained one copy of the red 3 ¶-ERG signal (''1 Edel'' and ''1 Esplit'') and those that exhibit more than one copy of the 3 ¶-ERG FISH signal (''2+ Edel'' and ''2+ Esplit'') (12).Class Normal tumors were classified as 3+ Normal if the presence of three or more copies of wild-type (WT) ERG were observed (Supplementary Table S1).For PTEN loss and ERG gene status, patients were classified into the potentially more aggressive class, i.e., homozygous versus heterozygous loss and 2+ Edel or 2+ Esplit versus Edel or Esplit.
FISH using an ERG break-apart assay was also performed, as described previously (12), on tissue slices adjacent to H&E sections confirmed to contain cancer.All the tissues were then scanned using an Ariol SL-50 scanner (Applied Imaging), and all areas of tumor were scored for ERG gene status.Tumors were classified by ERG gene class, as used for CTCs (Supplementary Table S1).

Results
Detecting CTCs in CRPC.Blood was collected from 89 CRPC patients being treated in phase I/II clinical trials of abiraterone acetate; 54 patients were chemotherapy-naive and 35 had received docetaxel.Overall, most patients had metastatic disease with only four chemotherapy-naive patients having an increasing PSA with no radiological evidence of disease on computed tomography scans or bone scan (Table 1).All patients had CK+ and CK+ CD45À events with relatively few being classified with an intact nucleus and cytoplasm (Supplementary Table S2).Of these intact CTCs, those that are undergoing apoptosis did not provide a FISH signal (Supplementary Table S3).A minimum of FISH results on four CTC was required to ensure that the result was technically reproducible, as the evaluation of FISH scores in 1,000 WBC controls revealed a 3.6% and 4.3% false-positive rate for the presence of ERG gene rearrangements and heterozygous PTEN loss, respectively (36/1000 WBC had splitting or loss of ERG locus probe and 42/1000 WBC had loss on one PTEN locus probe in a nucleus; Supplementary Table S4).Four or more CTC events were identified at any time point on the study in 51 patients, all of whom had radiologically detectable metastatic disease.The number of CTCs declined with treatment with abiraterone acetate (Supplementary Fig. S2; Supplementary Table S5) and was concordant in this cohort of patients with changes in PSA (correlation coefficient, r = 0.730; P < 0.0001, Spearman rank test; Supplementary Fig. S3).We confirmed, in this cohort of patients, previous observations that the number of CTCs is associated with survival and that a decline in the counts after treatment was associated with an improvement in outcome (Supplementary Figs.S4 and S5; refs. 15-17).  1 and Supplementary Table S6).Previous studies investigating TMPRSS2-ERG in CRPC used materials collected at rapid autopsy from a smaller number of patients and failed to detect any class Esplit tumors (18).In all our cases, no patient had four or more CTCs belonging to different ERG gene classes; no cases of CTCs with unrearranged ERG were observed in patients with an ERG rearrangement in biopsy tissue or CTCs.Also, in patients with an ERG rearrangement, no cases had four or more CTCs classified both class Edel and class Esplit.More than one copy of rearranged ERG was observed in 10 patients (2+ Edel or 2+ Esplit, 20% of all patients; Figs.1A and 2A), and 13 patients (26%) had CTCs with two or more copies of WT ERG (3+ Normal; Fig. 1B, 2B; ERG gene classes are explained in Supplementary Table S1).In all patients with more than two copies of the ERG gene locus, copy number was gained heterogeneously with single patients having CTCs with a range of different copy numbers (Fig. 2B).More than two copies of the ERG gene locus were associated with more than two copies All events had one copy of WT ERG, but event 476 has two copies of rearranged ERG , event 736 has one copy, and event 963 has three copies.Event 476 has three copies of AR and three of PTEN , event 736 has five copies of AR and two of PTEN , and event 936 has two copies of AR and two of PTEN .B, ERG WT ploidy.Patient 7 did not have rearrangement of ERG in any CTC.Event 135 has two copies of WT ERG and three copies of PTEN , event 267 has four copies of WT ERG and five of PTEN , and event 416 has six copies of WT ERG and eight copies of PTEN.The number of copies of AR does not exceed the number of copies of PTEN or WT ERG in any event shown, but seven CTC with more than five copies of AR but two of three copies of PTEN and WT ERG were observed (not shown).C, heterogeneous PTEN loss.Patient 26 is a case of homozygous PTEN loss as shown in events 236 and 588; the CTC in event 588 is adjacent to a control leukocyte with two copies of PTEN .There is loss of one copy of PTEN in event 13 and normal PTEN complement in event 505.ERG is not rearranged in any CTC; there are two copies of WT ERG in events 13 and 588, five copies in event 236, and three copies in event 505.There are seven copies of AR in event 13, three copies in event 236, one copy in event 505, and two copies in event 588; no other CTC had more than five AR copies.D, heterogeneous AR gain.Patient 71 has two or three copies of WT ERG and PTEN in all CTCs.Event 725 has one copy of AR , event 73 has two copies of AR , event 106 has six copies of AR , and event 423 has >10 copies of AR .
of PTEN in patients with no PTEN loss.In contrast, FISH studies revealed significant heterogeneity of PTEN loss and AR copy number gain.Thirteen patients (26%) had evidence of loss of PTEN in their CTCs (Table 1).Six of 13 patients had homogenous homozygous loss of PTEN in all their CTCs.Seven patients had heterogeneous PTEN loss: two patients each had four or more CTCs with homozygous and heterozygous loss and four or more cells with a full complement of PTEN (Fig. 1C), one patient had four or more CTCs harboring both heterozygous and homozygous PTEN loss, and four patients had four or more CTCs with both heterozygous loss and normal PTEN (Fig. 2C).All cases with heterozygous loss (n = 7) showed heterogeneity of PTEN loss but only three of nine with homozygous loss showed heterogeneity.AR gene copy number was available on CTCs for 33 patients: copy number gain was invariably heterogeneous (Fig. 1D).No patient had only one copy of AR in all their CTCs, whereas WBCs all revealed one AR copy.Twenty-eight patients (85%) had CTCs harboring a maximum of three copies of AR , 10 patients (30%) had CTCs with a maximum of four or five copies, and 15 patients (45%) had CTCs with more than five copies of AR (Supplementary Table S7).There was no change in ERG and PTEN status in CTCs collected from patients before starting and after progressing on abiraterone acetate (n = 24).Similarly, there was no shift in AR copy number with treatment with abiraterone acetate (n = 18; Supplementary Table S7).
ERG gene status does not change with the development of castration resistance.We then proceeded to compare the ERG gene status of CTCs with that of archival diagnostic tumor tissue acquired before the institution of first-line hormonal treatment (n = 27) and of CRPC biopsies (n = 11) from the same patient.We found that CTC ERG gene status matched the ERG gene status of tumor tissues in all cases (Table 2).These data also showed that all patients with a rearrangement of ERG in the therapy-naive tissue had ERG-rearranged cancer in the CRPC tissue; similarly, no CRPC tissue with ERG rearrangements was detected in patients who did not have a rearrangement at baseline (n = 31).One patient (patient 74) had class Normal and class Edel in his baseline therapy-naive samples; all his CTCs were class Edel.One patient (patient 70) had 70% class Edel and 30% class Esplit in his therapy-naive samples; his CTCs were class Edel.Similarly, a second patient (patient 2) had 65% class Esplit and 35% class Edel in his therapynaive samples; his CTC were class Esplit.
ERG remains expressed in CRPC.To confirm that ERG expression was maintained in CRPC samples with a TMPRSS2-ERG fusion detectable by FISH (break-apart assay), quantitative RT-PCR was performed for ERG exon 11 in CRPC tumor samples, and this was matched with therapy-naive samples.Expression of ERG exon 11 was significantly higher in all ERG-rearranged CRPC samples (n = 9) compared with CRPC samples with no ERG rearrangement (n = 6; P = 0.0004, Mann-Whitney test).Post-abiraterone acetate treatment biopsies were performed in three patients, but tumor tissue was identified in only one patient.A decrease in expression of ERG exon 11 was observed in this patient while responding to abiraterone acetate.Similarly, androgen-responsive expression of ERG exon 11 increased 3.5-fold in VCaP cell lines treated with R1881 (Fig. 3; Supplementary Table S8).
Clinical implications of assessing ERG gene status.As there was no change in ERG gene status with castration resistance or with abiraterone acetate treatment, we combined ERG gene FISH data from CTCs (n = 49), CRPC tumor tissue (n = 17), and archival, therapy-naive tissue (n = 53) to obtain an ERG gene class for 77 patients (we have been unable to acquire neither tumor tissue nor CTC for the remaining 12 patients).Thirty-two patients (41%) had an ERG rearrangement, and 45 did not.Nineteen patients were on a stable dose of steroids at the start of the study and continued the same steroid dose in combination with abiraterone acetate, and 58 patients were treated with single-agent abiraterone acetate.All patients were participating in phase I/phase II clinical trials with a primary efficacy end point of PSA decline as recommended by PSA working group I criteria (19).The presence of an ERG rearrangement was associated with the magnitude of maximal PSA decline on abiraterone acetate (P = 0.007, Mann-Whitney test for trend), with 12 of 15 patients (80%) who had a z90% PSA decline, having an ERG rearrangement compared with 20 of 62 patients (32%) who did not have a z90% PSA decline being ERG rearranged (P = 0.001, Fisher exact test).The association between an ERG rearrangement and declines in PSA of a smaller magnitude than 90% approached Novel antiandrogens that target the AR and small molecule inhibitors of the phosphatidylinositol-3-kinase (PI3K)/PTEN are being evaluated in clinical trials as a strategy for reversing resistance to hormone therapies (6).However, of the patients with evidence of PTEN loss, 7 of 13 had z50% declines in PSA and one (who had 100% homozygous PTEN loss and a class 2+ Edel ERG rearrangement) had a z90% PSA decline.Similarly, tumor responses were observed in patients with AR copy number gain.Moreover, loss of PTEN was not observed at the development of resistance to abiraterone acetate.As this study reports multiple subclasses for every gene locus, for example, homozygous PTEN loss in 100%, 80% to 100%, 60% to 80%, 40% to 60%, and<40% of CTCs, and ERG gene rearrangements by deletion or rearrangement and with or without duplication, large numbers of patients need to be evaluated to investigate the effect of specific genetic aberrations on clinical outcome.This will be pursued in an ongoing randomized double-blind phase III trial of abiraterone acetate and prednisone versus prednisone and placebo.The significant genetic heterogeneity observed at the PTEN and AR loci contrasts to ERG homogeneity and, together with the absence of a change in ERG gene status with the development of castration resistance, could suggest that rearrangement of ERG is an earlier event in prostate carcinogenesis than gain of AR and, in some cases, loss of PTEN.Moreover, CTC genetic heterogeneity for AR amplification and PTEN loss may have important implications for the development of therapeutics that target either the AR or PI3K/AKT pathways.Finally, these data strongly suggest that circulating, non-apoptotic nucleated, EPCAM+ CK+ CD45-cells isolated from CRPC patients are malignant in origin and confirm that hormone-regulated expression of ERG persists in CRPC.The costs of publication of this article were defrayed in part by the payment of page charges.This article must therefore be hereby marked advertisement in accordance with 18 U.S.C.Section 1734 solely to indicate this fact.

Disclosure of Potential Conflicts of Interest
We thank patients who participated in these studies and their families.

Figure 1 .
Figure 1.Molecular characterization of CTCs.CK (green ) and CD45 (red) immunofluorescence and PTEN, ERG break-apart, and AR FISH.Examples of events from four patients that were CK positive and had an intracellular 4 ¶,6-diamidino-2-phenylindole-stained nucleus are included.There are three examples of control leukocytes that stain positive for CD45 (red) and negative for CK (no green ; A, event 46; C, events 505 and 588).A, class Edel with ERG gene copy number heterogeneity.Patient 86 had loss of the 5 ¶-ERG signal suggesting fusion of TMPRSS2 with ERG secondary to deletion.All CTC analyzed were class Edel.All events had one copy of WT ERG, but event 476 has two copies of rearranged ERG , event 736 has one copy, and event 963 has three copies.Event 476 has three copies of AR and three of PTEN , event 736 has five copies of AR and two of PTEN , and event 936 has two copies of AR and two of PTEN .B, ERG WT ploidy.Patient 7 did not have rearrangement of ERG in any CTC.Event 135 has two copies of WT ERG and three copies of PTEN , event 267 has four copies of WT ERG and five of PTEN , and event 416 has six copies of WT ERG and eight copies of PTEN.The number of copies of AR does not exceed the number of copies of PTEN or WT ERG in any event shown, but seven CTC with more than five copies of AR but two of three copies of PTEN and WT ERG were observed (not shown).C, heterogeneous PTEN loss.Patient 26 is a case of homozygous PTEN loss as shown in events 236 and 588; the CTC in event 588 is adjacent to a control leukocyte with two copies of PTEN .There is loss of one copy of PTEN in event 13 and normal PTEN complement in event 505.ERG is not rearranged in any CTC; there are two copies of WT ERG in events 13 and 588, five copies in event 236, and three copies in event 505.There are seven copies of AR in event 13, three copies in event 236, one copy in event 505, and two copies in event 588; no other CTC had more than five AR copies.D, heterogeneous AR gain.Patient 71 has two or three copies of WT ERG and PTEN in all CTCs.Event 725 has one copy of AR , event 73 has two copies of AR , event 106 has six copies of AR , and event 423 has >10 copies of AR .

Figure 2 .
Figure 2. Homogeneity of ERG gene rearrangement and heterogeneity of PTEN loss and AR copy number gain.CTC with FISH abnormalities in three or fewer CTCs are not represented A, ERG gene copy number of all 23 patients with rearrangement.Fifteen patients are of class Edel, and eight are of class Esplit; eight cases had CTC harboring both 1 Edel and 2 Edel, and two cases had CTC harboring both 1 Esplit and 2+ Esplit.No patient had four or more CTC belonging to more than one ERG gene class.B, WT ERG gene ploidy (class 3+ Normal).Twelve patients had CTC harboring three or more copies of WT ERG.C, heterogeneity of PTEN loss.Thirteen patients had CTC with evidence of loss of PTEN.Six patients had homogenous homozygous loss of both PTEN loci.The other seven patients had CTC with heterogeneous loss of PTEN .
All patients participating in clinical trials were sponsored by Cougar Biotechnology.Processing of blood samples for CTC evaluation was conducted in a research collaboration with Immunicon Corporation.However, neither Cougar Biotechnology nor Immunicon Corporation had any input to the study design, analysis of samples, or interpretation of the data.J.F. Swennenhuis, R. Levink, F. Coumans, R. Sipkema, and L.W.M.M. Terstappen were employees of Immunicon Corporation, which has been purchased by Veridex LLC.Veridex LLC owns the rights to CTC evaluation using the technology described in this manuscript.L.W.M.M. Terstappen is a consultant for Veridex LLC. A. Molina is a full-time employee of Cougar Biotechnology, which holds the rights of development of abiraterone acetate.J.S. De Bono has served as unpaid consultant for Cougar Biotechnology.Abiraterone acetate was discovered at The Institute of Cancer Research, which therefore has a commercial interest in the development of this agent.G. Attard, D. Olmos, A.H.M. Reid, R. A'Hern, J. Moreira, R. Riisnaes, E. Thompson, C.P. Carden, C. Parker, D. Dearnaley, S.B. Kaye, C.S. Cooper, J.S. De Bono are employees of The Institute of Cancer Research.

Figure 3 .
Figure 3. ERG remains expressed in CRPC.The bar chart represents expression of tumor samples obtained from patients treated on clinical studies of abiraterone acetate, normalized to the expression of ERG by VCAP cells in normal serum.Columns in red on the left side of the chart represent patients with an ERG rearrangement confirmed by FISH and a TMPRSS2-ERG fusion transcript confirmed by RT-PCR.Columns in blue in the middle of the chart represent patients with no ERG rearrangement.Striped columns signify a therapy-naive sample matched to a castration-resistant sample.One patient had castration-resistant samples collected immediately before starting abiraterone acetate and a second set of samples acquired while receiving treatment with abiraterone acetate (column with diamond pattern ).The columns in green on the right side of the chart represent expression of ERG by VCaP cells in charcoal-stripped serum, FBS, and added synthetic androgens (R1881).Bars, SE.The maximal percentage of PSA decline, and time-on-treatment on abiraterone acetate for these patients is reported in Supplementary TableS8.TNPC, therapy-naive prostate cancer; AA, abiraterone acetate; CSS, charcoal-stripped serum; R1881, synthetic androgen added to serum.
G. Attard and J.S. De Bono designed the study, interpreted the data, and wrote the first draft of the manuscript.J.F. Swennenhuis, R. Levink, F. Coumans, R. Sipkema, and L.W.M.M. Terstappen developed the technology for CTC identification and FISH on CTC, performed FISH on all the CTCs, and assisted G. Attard with the analysis of the samples.E. Vickers and M.E.Cox assisted G. Attard with the experiments for analysis of ERG expression and, with G. Attard and J.S. De Bono, interpreted the results.G. Attard, J.S. De Bono, D. Olmos, A.H.M. Reid, N. Babu Oommen, C.P. Carden, C. Parker, D. Dearnaley, and S.B. Kaye contributed patients to this study.G. Hawche and E. Thompson assisted with collection of the samples.J. Moreira and C.S. Cooper assisted G. Attard with FISH on tissue samples.C. Jameson scored all samples for areas of cancer and contributed expert pathologic advice.R. Riisnaes sectioned all the tumor samples.A. Molina was involved in the administration of the abiraterone acetate studies on behalf of Cougar Biotechnology.R. A'Hern and D. Olmos gave expert statistical advice on interpreting the results.All authors contributed to the final version of the manuscript.

Table 1 .
Patient characteristics divided by prior chemotherapy treatment NOTE: ERG and PTEN CTC gene status was available for 49 of 51 patients (96%) with four or more CTCs at any time point on study.AR copy number was evaluated in 33 of 51 patients (64%) with four or more CTCs at any time point on study (maximum copy number observed is reported).

Table S8 .
TNPC, therapy-naive prostate cancer; AA, abiraterone acetate; CSS, charcoal-stripped serum; R1881, synthetic androgen added to serum.supported by the Prostate Cancer Foundation.D. Olmos was supported by an ESMO fellowship.C.S. Cooper was supported by the Grand Charity of Freemasons.C.S. Cooper and C. Parker were supported by the National Cancer Research Institute Prostate Cancer Collaborative.R. A'Hern was funded by Cancer Research UK Section of Clinical Trials.E. Vickers and M. Cox were supported by The Terry Fox Foundation and National Cancer Institute of Canada.We acknowledge NHS funding to the NIHR Biomedical Research Centre.

Table 3 .
ERG rearrangements are associated with magnitude of PSA decline on abiraterone acetate The number of patients with a PSA decline of z90% (I), z50% (II), and z50-89% and z90% (III) are divided by ERG gene status.